rsyslog/rsyslog
1
0
Fork 0
mirror of https://github.com/rsyslog/rsyslog.git synced 2025-12-18 01:40:42 +01:00
Code Issues Packages Projects Releases Wiki Activity

slightly improved lock contention situation by moving out of

Browse Source 
the critical section what could so with acceptable consequences
This commit is contained in:
Rainer Gerhards 2008-10-07 11:46:46 +02:00
parent 7b3c05da9f
commit cdecd7e524
1 changed files with 9 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
runtime/queue.c
View File

@ -2101,6 +2101,15 @@ queueEnqObj(queue_t *pThis, flowControl_t flowCtlType, void *pUsr)
ISOBJ_TYPE_assert(pThis, queue);
/* first check if we need to discard this message (which will cause CHKiRet() to exit)
* rgerhards, 2008-10-07: It is OK to do this outside of mutex protection. The iQueueSize
* and bRunsDA parameters may not reflect the correct settings here, but they are
* "good enough" in the sense that they can be used to drive the decision. Valgrind's
* threading tools may point this access to be an error, but this is done
* intentional. I do not see this causes problems to us.
*/
CHKiRet(queueChkDiscardMsg(pThis, pThis->iQueueSize, pThis->bRunsDA, pUsr));
/* Please note that this function is not cancel-safe and consequently
* sets the calling thread's cancelibility state to PTHREAD_CANCEL_DISABLE
* during its execution. If that is not done, race conditions occur if the
@ -2112,9 +2121,6 @@ queueEnqObj(queue_t *pThis, flowControl_t flowCtlType, void *pUsr)
d_pthread_mutex_lock(pThis->mut);
}
/* first check if we need to discard this message (which will cause CHKiRet() to exit) */
CHKiRet(queueChkDiscardMsg(pThis, pThis->iQueueSize, pThis->bRunsDA, pUsr));
/* then check if we need to add an assistance disk queue */
if(pThis->bIsDA)
CHKiRet(queueChkStrtDA(pThis));